It’s always nice to read good news. And it’s nice to read evidence-based arguments in the popular press. Over the holiday, Andrew Mack and Steven Pinker offered a little of each over the holidays in their article “The World Is Not Falling Apart.” Therein, they marshal of human security indicators upon indicators – number of rapes reported, number of civilians killed, number of wars breaking out, number of homicides – to argue that at the global level the trendlines are mostly pointing downward. In championing “an evidence-based mindset on the state of the world,” the authors place the blame for our current misconceptions on “a misleading formula of journalistic narration”:
“Reporters give lavish coverage to gun bursts, explosions and viral videos, oblivious to how representative they are and apparently innocent of the fact that many were contrived as journalistic bait… News is about things that happen, not things that don’t happen… The only sound way to appraise the state of the world is to count. How many violent acts has the world seen compared with the number of opportunities?”
This seemingly sensible argument does contain one fundamental paradox, however: some of the data-sets on which Mack and Pinker rely are themselves based on news reports. The Uppsala One-Sided Violence data-set, for example, which is their key source for the both statistic that mass killing has declined and the statistic on the decline of war, is based on keyword searches in the Factiva database of news reports.
If the news media tend to glamorize the worst events (as the authors point out), does this not imply that there may be some missing data? Megan Price and Patrick Ball have written about “event size bias” in journalism, for example: journalists are far likelier to report on massacres of many civilians than on the steady drip of civilian deaths that cumulate in smaller discrete numbers. Similarly, we are likelier to hear reports of dead women and children than dead men, and reports in urban areas rather than far away from news offices. One implication of Mack and Pinker’s argument may be that we need to look at trendlines based on estimates of total numbers, rather than counting only the things we can observe.